|Home | About | Journals | Submit | Contact Us | Français|
The incidence of melanoma in the US is increasing at a greater rate than that of any other cancer site (Rigel, 1997), and this is considered real and not because of earlier detection or changes in diagnostic criteria (Dennis, 1999). Excessive exposure to sunlight is the primary environmental risk factor for melanoma (Elwood and Jopson, 1997; Osterlind et al, 1988). Ultraviolet (UV) radiation from the sun produces free radicals that may initiate carcinogenesis through oxidative damage to DNA, and antioxidant nutrients, such as carotenoids and vitamins E and C, can neutralize the free radicals or even suppress their creation (Di Mascio et al, 1991; Frei, 1994). In animals, β-carotene and α-tocopherol can inhibit skin tumours induced by UV light (Black and Chan, 1975). Preformed vitamin A (i.e., retinol) is necessary for the regulation of cell differentiation and can also reduce the incidence of skin tumours in animals exposed to UV light (Levine and Meyskens, 1980; Oikarinen et al, 1991). Certain carotenoids (e.g., α-carotene, β-carotene, β-cryptoxanthin) are metabolised to retinol. These retinol precursors along with antioxidant nutrients are prominent components of fruits and vegetables and may account for the observed anticancer effects of these foods (Ziegler, 1991).
Although antioxidants and retinol appear to be beneficial for some cancers, little is known about their effect on the development of melanoma. In case–control studies, lower risks of melanoma have been associated with higher blood levels of α-tocopherol (Knekt et al, 1991), β-carotene (Knekt et al, 1991), and retinol (Breslow et al, 1995), and with greater dietary intakes of vitamin E (Stryker et al, 1990; Bain et al, 1993; Kirkpatrick et al, 1994) and β-carotene (Bain et al, 1993). Only one prospective study of diet and melanoma has been reported and no associations were found for carotenoids, retinoids, or vitamins C and E (Veierod et al, 1997), although the questionnaire was inadequate for accurate assessment of these nutrients.
In this prospective study within two cohorts of women, we examined the risk of melanoma by intakes of vitamin C, vitamin E and specific tocopherols, retinol, and β-carotene and other carotenoids.
The Nurses' Health Study (NHS) cohort consists of 121700 female registered nurses living within 11 US states who were 30–55 years of age when they responded to the initial questionnaire in 1976. A second cohort of younger women, NHS II, was begun in 1989 with 116671 nurses 25–42 years of age across 14 US states. Both studies were designed to examine prospectively the effects of lifestyle on risk of chronic disease. Information on individual characteristics, behaviours, and diagnosed diseases is collected with mailed questionnaires every 2 years and a response rate of at least 90% has been attained in each follow-up cycle. Deaths are confirmed through the National Death Index (Stampfer et al, 1984).
A baseline food frequency questionnaire (FFQ) was included in the NHS and NHS II mailings in 1980 and 1991, respectively, and diet has been reassessed at least every 4 years. In this analysis, we used the 1984 FFQ as the baseline measure for the NHS cohort since it contained 38 fruits and vegetables, significantly more than the initial 1980 questionnaire (Feskanich et al, 2000).
The NHS and NHS II populations for this analysis consisted of those participants who completed an FFQ at baseline and who had no previous report of cancer (except nonmelanoma skin cancer). Owing to small numbers and low risk, non-Caucasian women were also excluded. After exclusions, 73525 women remained in the NHS study population and 88553 women remained in the NHS II study population.
Within the study populations over the periods of follow-up for these analyses, 571 NHS women and 378 NHS II women reported a diagnosis of melanoma. Medical records were obtained for 486 (85%) of the NHS reports and for 289 (76%) of the NHS II reports; 254 (52%) and 160 (55%), respectively, were confirmed cases of invasive melanoma. These included superficial spreading and nodular types.
For each food on the FFQ, women reported their frequency of intake during the past year in terms of a specified serving size. Daily dietary intakes of nutrients were calculated based on the nutrient content of foods derived primarily from US Department of Agriculture sources and supplemented with data from food manufacturers and published research. Total daily nutrient intakes were calculated by adding the amounts from multivitamins and specific supplements to the intakes from food. Use of nutrient supplements was assessed on each biennial questionnaire. For multivitamins, participants provided the name brand and the number of tablets taken per week. For vitamins C, E, and A (retinol) supplements, participants specified their daily dosage. For β-carotene supplements, reported use was assumed to be 6000μgday−1.
Nutrient intakes were adjusted for total energy consumption (Willett and Stampfer, 1986) and these values were cumulatively updated in analyses to create the best estimates of long-term intake. That is, at the beginning of every 2-year follow-up cycle, each nutrient intake was calculated as the mean of all reported intakes up to that time.
Results from several validation studies in the NHS cohort indicate that nutrient intakes from the FFQ are valid for ranking subjects in diet analyses. In a comparison of the 1986 FFQ with two 1-week diet records, correlations were 0.79 for vitamin A from foods and 0.76 for vitamin C from foods (Willett, 1998). In other studies, vitamin E and β-carotene intakes from the FFQ predicted plasma levels of α-tocopherol (r=0.55) (Stryker et al, 1988) and β-carotene (r=0.31) (Michaud et al, 1998), respectively.
In the NHS cohort, a brief FFQ was included in the 1986 questionnaire for reporting food consumption during high school. These data were used to determine intakes earlier in life and to identify women with consistent intakes in high school and adult years.
In both cohorts, age, body weight, menopausal status, parity, and use of oral contraceptives or postmenopausal hormones were assessed in each 2-year follow-up cycle. Height was reported only on the initial questionnaire and was used to calculate body mass index (BMI, kgm−2) in each cycle. Statistical analyses were also controlled for the following factors: skin reaction after two or more hours in the sun during childhood; number of sunburns over the lifetime (NHS) or between ages 15 and 20 (NHS II); number of moles on the left arm (NHS) or lower legs (NHS II); natural hair colour at age 21; family history (parent or sibling) of melanoma; and state of residence at cohort initiation. Information on frequency of sunscreen use was also collected, but was not used in analyses because it was unrelated to dietary intakes and was not independently associated with a decreased risk of melanoma.
The NHS and NHS II cohorts were analysed separately. Person-time was accrued for each participant from the return date of the baseline questionnaire (mailed June 1984 for NHS; June 1991 for NHS II) until a confirmed diagnosis of melanoma, a report of another cancer, death, or end of follow-up (1 June 1998 for NHS; 1 June 1999 for NHS II). Total person-years in analysis included 956356 from NHS and 691784 from NHS II.
Current diet and covariate data were used to allocate person-time to the appropriate category for each variable at the beginning of each 2-year follow-up cycle. Age-adjusted incidence rates were calculated within categories of diet and supplement use, and relative risks (RRs) were then calculated as the ratio of risk in each upper category compared to the risk in the lowest, or referent, category. We used proportional hazards models to adjust the RRs for all covariate information. To examine linear trend, dietary intakes were entered into models as continuous values.
To increase precision and to obtain a single summary of the results from NHS and NHS II, a random effects model (DerSimmonian and Laird, 1986) was used to combine the RRs and to test for heterogeneity of results (P<0.05).
Age-specific incidences of melanoma in the study populations were similar to national rates. For example, for white women 50–54 years of age, rates were 25 out of 100000year−1 in NHS and 27 out of 100000year−1 in recent US data (Ries et al, 2002).
Table 1 shows the characteristics of the study populations within low and high categories of vitamin C, vitamin E, retinol, and β-carotene intakes from foods plus supplements. For all nutrients, NHS women with higher intakes were older, and if postmenopausal, were more likely to use postmenopausal hormones. NHS II women with higher intakes were more likely to be nulliparous. In both cohorts, the nondietary risk factors for melanoma were not associated with nutrient intakes. Use of multivitamins and vitamin C and E supplements were high in both cohorts (53, 35, 36%, respectively, for NHS in 1998; 51, 27, 21%, respectively, for NHS II in 1999). Vitamin A (<5%) and β-carotene (<3%) supplement use were low in both cohorts.
We found no evidence that higher intakes of vitamin C, vitamin E, retinol, or β-carotene, the major carotenoid with provitamin A activity, are associated with a lower risk of melanoma (Table 2 ). Specific tocopherols (α-tocopherol, β-tocopherol, δ-tocopherol, γ-tocopherol) and other carotenoids, with or without provitamin A activity (α-carotene, lycopene, β-cryptoxanthin, lutein, and zeaxanthin), were also unrelated to risk (data not shown). Age-adjusted relative risks were only modestly attenuated when adjusted for the other melanoma risk factors.
Rather than the hypothesised inverse relation, we observed a positive association between dietary vitamin C and melanoma. Women with intakes 175mgday−1 had an RR of 1.43 (95% CI 1.02–2.00) compared to those with intakes <90mgday−1. The dose – response association was also marginally significant (P=0.05). A significant linear relation remained when vitamin E, retinol, and β-carotene were added to the model (P=0.03).
We also examined nutrient supplements in relation to melanoma risk (Table 3 ). Current use of multivitamins (a major source of retinol), and supplements of vitamins C and E were all unassociated with risk, and there was also no indication of benefit with longer duration of use. Vitamin A and β-carotene supplement use were too low to provide sufficient power for analysis.
Total fruits and vegetables or those with a high carotenoid content (2000IUserving−1) were also not associated with a lower risk of melanoma (Table 4 ) (see Feskanich et al (2000) for a description of the fruits and vegetables). However, consumption of fruits and vegetables high in vitamin C (30mg serving−1) was significantly associated with an increasing risk (P for linear trend=0.01). Orange juice was the major contributor to dietary vitamin C (about 20% in both cohorts), and 1servingday−1 was associated with an elevated risk of melanoma when compared to no orange juice consumption (RR=1.61, 95% CI 0.92–2.84, P for linear trend=0.008). In the NHS cohort, women recalled their frequency of orange juice consumption during high school, and this too was associated with an increased risk of melanoma (RR= 1.59, 95% CI 1.17–2.18 for 1servingday−1 vs <1servingweek−1). When analysed in the same model, significant positive dose–response relations were observed for both adult (P=0.01) and high school (P=0.04) orange juice consumption. In an analysis of consistent intake over time, women who reported that they consumed 1servingday−1 of orange juice in both high school and as an adult had an RR of 2.32 (95% CI 1.36–4.00) compared to those who reported <1 serving week−1 in both time periods.
Since dietary vitamin C was not expected to be a risk factor for melanoma, we speculated that the positive associations might be because of greater sun exposure among women with higher vitamin C intakes. The highest incidences of melanoma were found among the NHS women who lived in Florida (43 out of 100000year−1) and California (31 out of 100000year−1), states where excessive sun exposure is more likely. However, their dietary intakes of vitamin C were not higher than the rest of the cohort and excluding them from analyses did not change results from those in Table 2 that were adjusted for state of residence.
Since nondietary determinants of melanoma contribute significantly to risk, we examined associations between vitamins A, C, E, and melanoma separately among women at higher and lower risk based upon nondietary factors. A total risk score was calculated using cohort-derived RRs associated with each of the five nondietary factors listed in Table 1. The lowest score of 5 was assigned to women with the lowest relative risk (RR=1.0) for all factors; the highest risk scores were 11.9 in NHS and 14.3 in NHS II. Categories of low, middle, and high risk were defined to ensure approximately equal numbers of cases in each category. Table 5 shows results for dietary vitamin C and total retinol, the only nutrients which exhibited possible interactions with the risk score. For dietary vitamin C, a significant positive association was found only among women in the high-risk category (P for linear trend=0.01), although the test for interaction was not significant (P=0.12). For total retinol, a significant inverse association was observed in the low-nondietary-risk category. An RR of 0.39 (95% CI 0.22–0.71) was associated with a total retinol intake of 1800μgday−1 when compared to <400μgday−1 (P for linear trend=0.01) and there was a significant interaction between total retinol and risk score (P=0.001).
In this prospective study within two large cohorts of Caucasian women 25–77 years of age, we found little evidence that vitamin C, vitamin E, individual tocopherols, retinol, β-carotene, or other carotenoids are associated with a lower risk of invasive melanoma. The only inverse association that we observed was for total retinol intake when limited to women who were otherwise at low risk of melanoma based on nondietary factors. Although not an antioxidant, retinol is essential for regulating cell differentiation, and it is possible that it may only exert a notable effect without the overwhelming influence of other major risk factors. There is little prior research to support our finding. Breslow et al (1995) reported an RR of 0.4 when comparing extreme tertiles of plasma retinol levels, although there were only 30 melanoma cases in this analysis and results were not significant. Other case–control studies did not find an association between plasma (Stryker et al, 1990; Knekt et al, 1991) or dietary (Stryker et al, 1990; Kirkpatrick et al, 1994) levels of retinol and risk of melanoma. It is possible that our observed inverse association for total retinol intake among the low-risk women is a chance finding, but even if real, we would not recommend high retinol intakes for women at such low risk of melanoma since higher intakes have been associated with increased risk of hip fracture among postmenopausal women (Feskanich et al, 2002).
An unexpected finding was the increased risk of melanoma with higher intakes of vitamin C from food, and more frequent consumption of fruits and vegetables high in vitamin C, particularly orange juice. It is possible that this, too, is a chance finding. However, given that this association was strongest among the higher risk sun-sensitive women, it is also possible that a photosensitising compound in these foods is responsible for the increased risk; vitamin C is unlikely to be responsible because supplement users were not at higher risk. Furocoumarins are naturally produced in fruits and vegetables in response to physical damage, fungal infection, and attack by pests, and they exhibit phototoxic effects because of their capacity to absorb photons and then release the energy to cause cellular damage (Gasparro et al, 1997). Plants of the umbelliferae family, such as parsnips, celery, fennel, and parsley, have a high furocoumarin content, and dermatologic reactions have been observed with ingestion of large quantities and subsequent exposure to UV radiation (Ljunggren, 1990; Puig and Demaragas, 1994). Some types of limes (Nigg et al, 1993) and oranges (Joint Food Safety and Standards Group, Food Standard Agency, UK, 1993) also contain appreciable quantities of furocoumarins and have been noted to trigger dermatologic reactions when accompanied by sun exposure (Gross et al, 1987; Wagner et al, 2002). Psoralen is a furocoumarin that has long been used with UV A radiation to treat psoriasis and other skin conditions (Parrish et al, 1974) and more recently has been associated with higher risk of squamous cell cancer (Lever and Farr, 1994; Stern et al, 1998) and melanoma (Stern et al, 1997). Whether the amounts of furocoumarins in orange juice, when ingested over many years, are sufficient to affect risk of melanoma in susceptible individuals is unclear.
The major strength of this study is the prospective design; previous research has been limited to retrospective studies. Other strengths are the wide age distribution of women, the repeated measures of diet and supplement use, and the ability to control for several major risk factors for melanoma. Although we lacked lifetime information on time spent in the sun, the risk factor data that we did collect were unrelated to diet and therefore did not confound results.
In conclusion, higher intakes of vitamins C, E, and A were generally not associated with lower risks of melanoma in Caucasian women, although retinol may be protective among women who are otherwise at low risk of disease. Certain foods, such as orange juice, may be associated with an increased risk of melanoma, but further research is needed to confirm this finding and to identify the responsible constituents in these foods.
We thank Dr Lois Gold for information on non-nutritional constituents in orange juice. Supported by research Grants CA87969, CA50385, and CA93683 from the US National Institutes of Health.